Tools

Open evaluation frameworks for safe software development. Evaluate AI infrastructure, scan for anti-patterns, and make informed release decisions.

The constraint-mandatory approach

Every tool on this page applies the same principle: safety should be the easiest path, not the hardest. When the architecture does the work, everyone who builds on it benefits.

We build evaluation frameworks alongside the infrastructure they test. These tools came out of building protocols and standards for the next generation of AI products, and they exist to help developers, researchers, and organisations make informed decisions about safety, privacy architecture, and responsible release.